Has LiDAR Arrived?

The Consumer Electronics Show (CES) is the annual mega-event held in the first week of the New Year in Las Vegas, Nevada (maybe not as mega this year for obvious reasons). It reveals future trends in consumer applications ranging from mobile communications, smart homes and healthcare to transportation, robotics and gaming. Over the past five years, automotive and transportation companies have increased their presence dramatically - to display what the future will bring in terms of energy efficiency, electrification, autonomy, comfort features, entertainment, new mobility modes, safety and styling. Given the increased importance of safety in autonomy, LiDAR companies have followed suit.

Is LiDAR technology moving away from hype, hopes and promises towards real applications and profitable products? CES and other recent events indicate that it is, for automotive and more broadly for AoT™ (Autonomy of Things), as well as smart city/infrastructure and security applications.

The major activities/trends occurring currently in the LiDARVerse include:

1) Announcements of automotive design wins, primarily for ADAS (Advanced Driver Assistance Systems) as well as new product launches.

2) A focus on more immediate, non-automotive applications like robotics, smart cities, security and logistics. The increased importance of 3D image processing and actionable perception software is becoming critical for these applications.

3) Solutions that use classical machine vision-based techniques for 3D imaging and depth mapping. If these succeed, the business case for automotive LiDAR gets substantially disrupted.

1) Public Announcements of Automotive Design Wins/New Designs

A major announcement was made by Valeo which secured a design win for its second generation SCALA LiDAR to support L3 functionality in the Mercedes S Class and Honda Legend (starting in 2022). Valeo also announced its third generation SCALA development to support L3 functionality at higher speeds (expected release in 2024). Currently, Valeo is the most experienced manufacturer of LiDAR for automotive applications in terms of units shipped, design wins and certification maturity.

According to Luminar’s recent S3 filing, “Daimler will share certain data from development and production vehicles with our lidars to be used for continuous product improvement and updates. In exchange for certain services provided under the strategic collaboration, we agreed to issue 1,500,000 shares of our Class A common stock to Daimler in a private placement transaction”. The collaboration is with Daimler North America and although portrayed otherwise by multiple tech and business media, is far from a design win or a hard financial investment at this point. It did wonders for the stock price however. Luminar also announced design wins in Volvo for L3 features and with trucking automation company TuSimple.

Cepton announced four months ago that its MMT (Micro-Motion Technology) based solid-state LiDAR had been designed into 9 different General Motors models with Super Cruise functionality. A more recent announcement indicates that this LiDAR will be deployed on the premium version of Super Cruise (Ultra Cruise) in 2023. Cepton is expected to go public in early 2022 via a SPAC transaction.

Other recent design win announcements include Innoviz (BMW 7 Series, with their MEMs based LiDAR), Innovusion (Nio Motors), Ibeo (ZF and Great Wall Motors), Hesai (Baidu, AutoX, Didi, Pony - uses surround view LiDAR for robotaxi applications) and Robosense (BYD, GAC).

New product launches were also announced at CES. Valeo announced its third generation SCALA as well as a short range LiDAR product. Hesai launched a 128 beam, opto-mechanically scanned LiDAR using VCSELs (Vertical Cavity Surface Illuminating Lasers) from Lumentum. Innoviz moved in an opposite direction - from its flagship MEMs scanned, long range solution t0 a 360° surround view LiDAR.

Opsys announced major customer collaborations at CES, including with South Korea based Tier 1 supplier, SL Corporation, to supply LiDAR units in 2024-2025. Opsys expects this to bring market penetration with Asian automotive OEMs. The company also launched its LiDAR building block product, the SP3. The unit was demonstrated in bright sunlight at at CES on a moving vehicle.

Some new players are also emerging in the automotive LiDAR space. These include:

• Preact, based in Portland, Oregon is a spin-out from ARTIS which made Active Protection Systems (APS) like the Iron Curtain for military vehicles. Preact’s TrueSense flash LiDAR camera uses high resolution ToF CCD (Time-of-Flight Charged Coupled Devices) arrays and LED (Light Emitting Devices) at 9XX nm. The focus is on short range (20 m) applications like curb detection, self-parking and imminent pedestrian detection. According to CEO Paul Drysch, a low cost point ($25) is key. State Farm led their Series A investment.
• Kyocera SLD, based in Santa Barbara, California focuses on high power, high brightness lasers in the visible wavelengths, primarily for illumination. SLD’s co-founders include Dr. Shuji Nakamura, a 2014 Nobel Laureate. The company was started in 2013 to focus on GaN based bright laser source, and acquired by Kyocera (a ceramics and materials powerhouse) in 2020. The LaserLight lasers are designed as illumination sources into headlamp assemblies for BMW models. The company demonstrated a 9XX version of this laser at CES in for a long range LiDAR. SLD CMO, Paul Rudy, believes that its laser solution enables 3D imaging at overall lower system cost and with industry proven components.
• Intel-Mobileye announced that it will launch a sub-$1000 coherent LiDAR based on its silicon photonics platform by 2024. This is an interesting development to watch as they position themselves to compete with Tesla (no LiDAR!) on the AV front. Mobileye also proposed a strategy of collecting enough safety data for AV operations by using one fleet equipped only with cameras and a second with radar and LiDAR. According to Mobileye, one million hours of driven miles can be extrapolated by driving each car for 10,000 miles (presumably, this is based on no accidents occurring, data sets being uncorrelated and the results applicable only to regions and conditions in which these miles are driven accident free). It is unclear why they would not equip a single fleet with all sensors and use only a subset to prove their point.
• Fastree-3D (Disclosure: I am an advisor), a developer of VCSEL and SPAD (Single Photon Avalanche Photodiodes) announced development contracts with a major European OEM and Tier 1 for its flash LiDARs with extremely low latency, integrated on- chip processing and sub-$100 cost. The goal is to service applications in industrial sensing (automated guided vehicles, robotics) and pedestrian detection (in automotive). The company had disclosed development contracts with Bosch and NXP last year.

2) Non-Automotive Applications and Importance of 3D Image Processing

All LiDAR companies want to play in the automotive space. It is the biggest prize in the battle to succeed and survive in the crowded LiDAR landscape. But it is difficult to be an automotive supplier. ADAS demands scale, reliability, quality, processes, culture, patience and a passion for reducing product costs which only Tier 1 automotive suppliers have the capabilities for today. There are a limited number of these not developing their own LiDAR (maybe 5?). The autonomous vehicle (AV) space is different - lower focus on costs, scale, quality and reliability. However, most highly funded AV companies like Argo, Aurora, Waymo and Yandex have internal LiDAR efforts because the software stacks are intimately linked to the LiDAR personality.

Going after non-automotive applications therefore becomes necessary for many LiDAR companies. While almost all say that they are addressing non-automotive applications (until automotive opportunities develop), some are more specific in these pursuits. The issue always is one of application specific customization, market size and justifying the billion dollar valuations of the companies that are public or want to be.

Quanergy, Ouster, Cepton, Velodyne and Oyla have engaged in efforts for LiDAR applications in non-automotive AoT™ and smart city/infrastructure and security applications.

• Quanergy was traditionally focused on automotive LiDAR (and in fact made LiDAR sexy on Wall Street) until a CEO change in 2020 pivoted the company into security, logistics and people counting applications. Interestingly, at CES 2022, they chose to exhibit in the Smart Cities section of the exhibition versus the predominant presence of other LiDAR companies in the Transportation section. They have made inroads into port automation and smart city applications with their surround view LiDAR. The maturity of the fully solid-state LiDAR using OPAs (Optical Phased Arrays) for AV applications is unclear, and was not displayed in detail at CES. Quanergy is planning to go public in 2022 via a SPAC transaction.
• Velodyne was the first pure-play LiDAR company to go public in 2020. It was also one of the pioneers of automotive LiDAR. High expectations of their dominance have largely evaporated after 2020, with a slew of management drama, and a recent filing of a lawsuit by David Hall, the founder. A new CEO was installed recently (Ted Tewksbury). One of his priorities is to ensure that new product designs are manufacturable, reliable and profitable. The company is also focusing on non-automotive applications like smart cities, smart infrastructure, mining and agriculture and claims a diversified customer base of ~300 customers. Integration of perception software into the LiDAR for certain applications is also an area of focus.
• Lumotive recently recruited a new CEO (Sam Heidari) and has increased its focus on industrial applications. The ability to scan laser energy using a solid state, metamaterials-based platform enables their LiDAR to be software controllable in terms of range performance in specific regions of interest. They recently proposed a universal API (Application Protocol Interface) to enable customers and software companies specializing in 3D imaging to optimize the LiDAR hardware for specific application requirements. Seoul Robotics, Cron.ai and Lake Fusion Technologies are their partners at this time. Lumotive recently won a Prism Award (at Photonics West in San Francisco).
• Ouster announced a cooperation with Serve Robotics which is a pioneer in autonomous package delivery. This builds on Ouster’s commitment to serve non-automotive applications in diverse areas like truckyard automation, logistics, mapping and mining. Customer wins with trucking automation and other robotics companies for logistics and mapping were also announced. Ouster recently acquired Sense Photonics to focus on automotive applications. The success of this acquisition and its integration, and penetration of the automotive market will be interesting to watch.
• Oyla: builds full stack solutions that can be mass deployed into broader markets like security (Disclosure: I am an advisor to Oyla). The solution fuses lidar, video and AI (Artificial Intelligence) in one edge device. Oyla’s uses high volume consumer components, and innovates on system, software and AI to build the LiDAR, and fuse it with color video. Srinath Kolluri, Oyla’s CEO: “Customers in non-automotive applications do not know how to make raw LiDAR data actionable. We remove a major pain point for these customers by our fusion-at-the-lens approach that pipes fused data into an end-to-end AI analytics engine.”
• Hesai and Innoviz were discussed earlier in this article. Their new product launches seem to be a pivot from their automotive focus, while holding on to their automotive socket wins. This is a challenging proposition because of few synergies with their existing product lines and a strong competitive landscape. It will be interesting to see how these develop.
• Seoul Robotics announced a major collaboration with BMW for logistics automation of vehicles rolling off the production line into holding parking lots. The company specializes in 3D image processing and used a networked system of hundreds of LiDAR units to guide autonomous driving and positioning of these vehicles. The role of specialized software providers for LIDAR is growing, especially for non-automotive applications. Companies like Outsight and Veuron are also active in this area, and had an active presence at CES.

3) LiDAR Disruptors

There are significant ongoing efforts that argue that LiDAR is an unnecessary crutch. Tesla is a classical example. The company has consistently argued that visible camera information from billions of travelled miles (in ~500M of its consumer cars driving worldwide) coupled with machine learning and artificial intelligence (AI) can be used to make ADAS and AVs practical. Sadly, promises of FSD (Full Self Driving) capability on its cars occurring within a year have been occurring regularly for the past five years, and customers are still waiting. Patiently. However, if Tesla and non-LiDAR approaches succeed, it will disrupt a multi-billion dollar ecosystem built around LiDAR, and more broadly the AV transportation space.

Some other companies working along the no-LiDAR angle include:

• Helm.ai has been developing camera and machine vision based systems for the past 5 years, primarily for high end ADAS and L4 AVs. As opposed to supervised deep learning approaches, Helm espouses unsupervised learning or deep teaching, wherein large amounts of data can used to train computers without image annotation and human intervention. Its ADAS products (highway and urban) use a combination of visible cameras and radar, whereas L4 products integrate LiDAR as well. It is not clear whether the LiDAR is currently installed to calibrate or compare information from the cameras and radar, and will eventually be removed. Helm is based in California, and recently raised $30M from Honda.
• Nodar (No LiDAR?), based in Massachusetts, uses stereo cameras to extract depth information. This is an established and mature discipline for short ranges. Nodar’s twist is that it can achieve long ranges (1 km) reliably through a combination of physical models and AI to filter out vibration effects that occur when the mounting distance between the cameras is increased (higher distance allows more range, but is also more susceptible to errors causes by manufacturing tolerance, road vibrations and changes in relative position over weather and time). Nodar is a CES® 2022 Innovation Awards Honoree and provided product demonstrations at the exhibition.
• Phiar is a Redwood City, California company focusing on making human driving safer. Their solution uses monocular cameras, GPS and maps, along with simulations and AI algorithms. The company announced a significant partnership with Qualcomm and Panasonic Automotive at CES. Phiar’s vision-based spatial AI technology will be used to support intelligent Augmented Reality (AR) heads-up display (HUD) for navigation and situational awareness. This provides drivers with significantly enhanced road perception and driving guidance. The company is secretive about how exactly it extracts depth information from monocular cameras. State Farm is an investor (for obvious reasons). Although this is currently being promoted as a Human in the Loop (HITL) driver assistance capability, it is not far-fetched to see the possibilities for higher levels of automation.

LiDAR has arrived, but the dice are still rolling in terms of the winners, key applications and time-frames. The possibility that LiDAR will be rendered irrelevant for automotive applications is ever present. Although unlikely, it could happen. What better place than CES and Las Vegas to see this fascinating landscape evolving?